If it is concern about how the MBS is interpreted causing worry, software vendors say, the department should look closer to home.
The fact that the Medicare Benefits Schedule is “basically impossible” for clinicians to use appropriately should be seen as a point in favour – not against – the existence of AI-powered billing tools, one software vendor tells The Medical Republic.
Last month, the Australian National Audit Office released a report looking at AI and Medicare benefits integrity.
Besides revealing that the Department of Health, Disability and Ageing had been using an AI-powered fraud detection system without having done a full risk assessment, the report also identified health provider billing practices as a potential threat.
According to the audit office, both Services Australia and the DoHDA had expressed concern about clinicians using AI tools to help identify which MBS items to bill.
The DoHDA also admitted that it had no way to tell whether a Medicare claim had been made using an AI billing assistant.
“At a high level, [DoHDA] and other Australian government entities have identified risks associated with AI use in healthcare settings, including in health provider billing,” the audit said.
“These AI risks include: potential for bias; erroneous outputs; patient data exposure; low practitioner AI literacy; insufficient performance monitoring; and patient consent issues.
“Services Australia has expressed concern to [DoHDA] about providers using AI to test MBS eligibility for multiple items and suggest billing options.”
Queensland GP Dr Patrick Gough, the co-founder of MBS interpreting software MediBetter, told TMR that the MBS, in its current form, was almost impossible for doctors to navigate.
“As everyone knows, you can email AskMBS and you’ll get four different answers, or you can call them and get a whole bunch of different answers … then if it came to legal proceedings you can’t even rely on that as defence,” he said.
“It’s always been a bit of a joke. The AI services that you have to provide basically have to work within that framework of ‘the MBS doesn’t know what it’s doing either’.
“We essentially then just provide options as to what [item numbers] could potentially be there, but then we have to then say you need to double check everything with AskMBS.”
Related
While there are more than 1000 items on the MBS that GPs are eligible to bill, Dr Gough said, GPs tend to only bill 20 to 30 individual items in their working life.
“There’s a lot of fear there, which is fine, which is understandable, because of the variation,” he said,
“But then there’s also just a lot of item numbers that they just have zero idea even exist, and there’s a lot of ones as well where they know it exists, but in their workflows they just miss what they’re doing.”
GP Dr Casey Going, co-founder of AI scribe and billing platform MBSPro, said products like MediBetter and MBSPro only gave suggestions and that it was up to the provider to decide whether to follow them.
One example would be an AI billing assistant listening into what a GP might have ordinarily billed as a regular consult and identifying that a patient mentioned their pain had been ongoing for more than six months, making them eligible for a chronic condition care plan.
Billing this item would likely come with a bigger rebate for the GP, making it more expensive for the government to support – even if it is appropriate.
Dr Going said he could see scenarios in which the number of GPs billing for care plans skyrocketed, causing concern for the department.
“AI prompting software is sounds like it’s revenue based, because we’re providing identification of care plans, but the idea is that we’re also providing education to the GP,” Dr Going told TMR.
“It’s not like ‘here’s a care plan to do 100 care plans’, it’s like ‘here’s a care plan, the patient has mentioned a little chronic disease and met some other criteria, and you should book your patient back in for review because that’s good medicine’.
“No one teaches us the MBS in medical school, so I think it’s filling in a gap that we probably have from the government where there hasn’t been a lot of education before or a lot of understanding of how these things work.
“Now we’ve got AI that can interpret some pretty complex rules contextually to the patient based on previous billing history, based on what they’ve said in the context, and all of a sudden that’s making a difference in terms of revenue outcomes, but also longitudinal care, where we’re seeing patients more regularly and more appropriately for the specific things that they need and billing those specific needs.”
